Torque-limited RecA polymerization on dsDNA Abstract The assembly of RecA onto a torsionally constrained double-stranded DNA molecule was followed in real time using magnetic tweezers. Formation of a RecA–DNA filament on the DNA tether was stalled owing to different physical processes depending on the applied stretching force. For forces up to 3.6 pN, the reaction stalled owing to the formation of positive plectonemes in the remaining DNA molecule. Release of these plectonemes by rotation of the magnets led to full coverage of the DNA molecule by RecA. At stretching forces larger than 3.6 pN, the twist induced during filament formation caused the reaction to stall before positive supercoils were generated. We deduce a maximum built-up torsion of 10.1 ± 0.7 kbT. In vivo this built-up torsion may be used to favor regression of a stalled replication fork or to free the chromosomal DNA in E.coli from its condensing proteins. INTRODUCTION In Escherichia coli, the RecA protein plays a central role in repair of damaged DNA through homologous recombination by mediating homologous DNA pairing and DNA strand-exchange reactions, in which a single-stranded DNA (ssDNA) replaces the homologous strand on a double-stranded DNA (dsDNA) (1). The structural modification of DNA by RecA during this genetic process is a key step in sequence recognition and specificity (2). Biochemical and electron microscopy studies have shown that RecA binds cooperatively to ss- and dsDNA with a stoichiometry of 1 RecA molecule per 3 nt or base pairs, respectively (3). The resulting complex is stretched by a factor of 1.5 with respect to B-form DNA and has a twist angle of 20° per base pair (4) instead of 35° per base pair in bare dsDNA. In B-form DNA, a single helical pitch measures 3.5 nm occupying 10.4 bases (5), whereas in a RecA–DNA filament this single helical pitch changes to 9.5 nm occupying 18.6 bases (4). Owing to these different helical pitches the binding of RecA causes unwinding of dsDNA. Electron microscopic images of RecA–DNA complexes suggest that RecA is unable to fully cover covalently closed circular DNA (4). Instead, it forms a partial filament, leaving the remaining uncovered DNA in a highly supercoiled state. The production of additional negative twist in the RecA–DNA filament is thus countered by the formation of positive writhe in the remaining DNA molecule. In contrast, open circular DNA can be fully covered by RecA. In vivo, however, the topology of dsDNA molecules is usually constrained. A torsional study allows us to explore the energetics involved in the assembly of RecA on topologically constrained DNA. In this paper, we used magnetic tweezers to study the dynamics of assembly of a RecA–DNA filament on a torsionally constrained DNA molecule in real time. As RecA changes the helical pitch upon binding, it generates torsion in the remainder of the DNA molecule. Depending on the applied stretching force, this generated torsion is stored either as writhe, in plectonemes, or as twist. In both the cases, RecA was found to form only partial filaments on the DNA molecule. Controlled release of the superhelical torsion resulted in a fully covered DNA molecule. From the torsion exerted by RecA on dsDNA (up to 10.1 kbT), an estimate of 7 kbT is deduced for the binding energy of a RecA monomer to dsDNA. Fulconis et al. (6) have recently reported a magnetic-tweezers study, in which they study the disassembly of a RecA–DNA filament in the presence of an external torque. They demonstrated that a DNA molecule can be fully covered by RecA when negatively unwinding a torsionally constrained DNA molecule prior to the introduction of RecA. MATERIALS AND METHODS Magnetic tweezers In the magnetic-tweezers setup used in these experiments (see Figure 1a), a DNA molecule is attached at one end to a glass surface and at the other end to a superparamagnetic bead (7). The tethered molecule can consequently be stretched and supercoiled by translating or rotating a pair of external magnets, respectively. The applied force is calculated by quantifying the thermal motion of the DNA-tethered bead (8). By using image processing, 5 nm position accuracy of the bead was obtained in three dimensions. To minimize the effect of thermal drift, all positions were measured relative to polystyrene beads that were fixed to the bottom of the flow cell. DNA substrates For assays involving rotationally constrained DNA, pSFV1 (Invitrogen) was cleaved with SpeI and BamHI, resulting in a 10 kb linear fragment, which was ligated to two 700 bp PCR fragments, one containing several biotin-modified dUTP bases and the other PCR fragment containing several digoxigenin-modified dUTP bases. Using the magnetic tweezers, we selected molecules that were rotationally constrained. As shown in Figure 1b, for forces below 0.5 pN, a torsionally constrained DNA molecule shows a symmetrical behavior upon rotation of the magnets. However, at higher forces the DNA tether shows an asymmetric behavior for positive and negative rotations by the magnets (e.g. the curve at F = 3.8 pN in Figure 1b), as described by Strick et al. (8). For experiments involving freely rotatable DNA, pSFV1 was cleaved with SpeI and XhoI, resulting in an 8 kb linear fragment. The biotin-modified PCR fragment was pre-treated with SAP to dephosphorylate the 5′ end. After ligation, the resulting construct therefore has a nick near the magnetic bead. Flow cell Polystyrene beads, as well as DNA constructs tethered to a magnetic bead at one end, were anchored to the bottom of a flow cell as described previously (7). The force-extension curve of a single DNA molecule was measured. After confirmation of the correct contour and persistence lengths, experiments were started with RecA. All measurements were carried out at 25°C. RecA/DNA reactions The flow cell final volume is ∼100 μl. All reactions were carried out in a buffer of 20 mM MOPS, pH 6.2, 10 mM MgCl2, 0.5 mM EDTA and 1 mM DTT. For the start of a measurement, the cell contents were replaced by 1 μM RecA (Roche) and 1 mM ATP in the same buffer, while the DNA molecule was relaxed. RecA polymerization in saturating conditions was monitored through measurement of the height of the magnetic bead, which measures the end-to-end distance of the DNA molecule. RESULTS Torsionally unconstrained dsDNA For nicked dsDNA, we observe the expected RecA-induced length extension of 50% (9–12). After the introduction of RecA and ATP in the flow cell, the height of the bead attached to a nicked DNA molecule was initially unaffected. However, after a short time delay (∼100 s), RecA assembly commenced, as observed from the extension of the DNA molecule (see Figure 2a). This was followed by a steady linear increase in the height of the bead. Finally, this increase ceased and the height of the bead remained constant. Stretching the DNA tether at different forces (0.5 < F < 9.0 pN) did not change the behavior of RecA polymerization on dsDNA (data not shown). The growth velocity appeared constant at ∼10 nm/s, or 20 monomers/s. The linear increase in height indicates a cooperative growth process. The observed growth profile is consistent with optical tweezer experiments by Shivashankar et al. (10), who stated that RecA assembly starts from a single nucleation site and proceeds linearly in one direction. The construct used for our experiment contained a nick, which is a preferable binding site to start nucleation, close to the magnetic bead. The change in contour length induced by RecA polymerization was 49 ± 3%, which is in agreement with results reported in literature (9–12). Torsionally constrained dsDNA In the case of torsionally constrained dsDNA, we observed a more complex behavior showing both elongation and shortening of the DNA tether, dependent on the applied force. At forces exceeding 3.6 pN, introduction of RecA into the flow cell initially showed the same behavior as torsionally unconstrained DNA (see Figure 2b). After a time lag, a linear increase in the height of the bead was observed with the same rate as above. The observed time lag was 2.9 ± 0.6 min (N = 14), allowing us to determine a nucleation rate of RecA onto torsionally constrained dsDNA of 3.4 ± 0.7 × 10−5 min−1 bp−1, which is in agreement with the results reported in literature (13). The DNA tether, however, was not extended by 50%, but by a much smaller value of 6.5 ± 1.1% (N = 8), independent of force between 3.6 and 10 pN. Apparently, at these forces RecA was not able to form a complete filament on the torsionally constrained dsDNA molecule. The height increase of 0.18 ± 0.03 μm (N = 8) implied that a RecA–DNA filament with a length of only 0.36 + 0.18 = 0.54 μm formed on the DNA molecule, i.e. a 0.36 μm fragment of DNA, was extended by 50%. The key difference with the previous experiment was that a torsionally constrained DNA molecule accumulates twist, which it could not release. During filament assembly, the binding of each RecA monomer induced a small twist in the remaining dsDNA. Therefore, as the filament grew in length, the twist stored in the remaining dsDNA molecule increased. After some point, the accumulated twist prevented the next RecA monomer from assembling, and extension of the dsDNA molecule was therefore stalled. A strikingly different behavior was observed for forces below 3.6 pN. In these cases, an initial increase in the height of the bead was followed by a large decrease (Figure 2c). At very low forces (below 0.5 pN), the molecule's end-to-end distance did not even show the initial increase but directly started to reduce. The process reached an equilibrium where no further shortening was observed (e.g. near 470 s in Figure 2c). The end-to-end distance of the molecule at which the reaction finally stalled varied strongly with different stretching forces. In Figure 3, these stalling heights are plotted versus force. These stalling heights correspond to the length of a partially RecA-coated dsDNA molecule as will be shown below. As can be seen, the stalling height increased monotonically with the applied stretching force up to 3.6 pN. However, further RecA assembly can be induced at the stalling position by negative rotation of the magnets (see Figure 4) that released (some of) the positive supercoils in the remaining uncovered DNA molecule. At constant stretching force, the end-to-end distance of the molecule increased with increasing negative rotation of the magnets to a final elongation of ∼50% (compared with the original DNA contour length). Beyond this point, the molecule's extension started to decrease owing to twist-induced RecA depolymerization as also shown by Fulconis et al. (6). In contrast, the application of positive instead of negative rotation of the magnets at the stalling position caused a decrease in the end-to-end distance of the molecule (data not shown). DATA ANALYSIS AND DISCUSSION When a RecA monomer binds to dsDNA, the base-to-base distance increases by 50% and the helix is unwound by 15° per base pair, because the twist angle per base pair changes from 35° to 20°. In the case of a torsionally unconstrained DNA molecule, this results in elongation because the torsion is released by free rotation of one DNA strand. However, when the DNA molecule is not allowed to freely rotate around its helical axis, the torsional energy is stored as either twist or writhe depending on the applied stretching force. At high forces, the DNA extension initially remains unchanged upon rotation (see the data for 3.8 pN in Figure 1b) and the torsional energy is stored as twist. At low stretching forces (see the data for 0.4 pN in Figure 1b), this torsional energy is stored as writhe (plectonemes) and the extension of the DNA molecule decreases upon rotation, because the plectonemes consume a significant length of DNA. This behavior explains why, in the low-force regime (0.5 < F < 3.5 pN), both elongation and shortening are observed during RecA assembly on torsionally constrained DNA. First, elongation is observed because RecA polymerization proceeds and the induced torsional energy is initially consumed as twist. When the DNA molecule reaches its buckling instability, twist cannot be increased further and additional torsional energy is consumed as writhe causing a reduction in the bead height owing to the formation of plectonemes in the DNA molecule. This process continues until the DNA molecule is fully consumed in one part that is covered by RecA and another part that is plectonemically supercoiled. At very low forces (F < 0.5 pN), the net effect is that the height of the bead decreases directly, because the length reduction caused by plectoneme formation dominates. Depending on the applied stretching force, the induced torsional energy owing to RecA polymerization is stored in twist or writhe (plectonemes), because the linking number (Lk) of a torsionally constrained construct remains constant (14). The linking number can either be increased or decreased by applying positive or negative magnet turns, respectively. Because a single turn of the magnets implies one turn on the molecule, we have ΔLk = r, where r is the number of turns the magnet rotates. As shown in Figure 4, the molecule reached its maximum length after 355 ± 10 negative rotations of the magnets at an applied stretching force of 1.8 pN, whereas 320 ± 10 rotations were needed to reach the maximum length at F = 0.4 pN. From force-extension measurements in both the cases, we obtained a contour length of 4.10 ± 0.05 μm. Prior to the introduction of RecA, the contour length of the relaxed DNA molecule was 2.72 ± 0.02 μm. Therefore, the effective length increase is 51 ± 2%. The linking number of a helical molecule is obtained by dividing its contour length by its helical pitch. In the case of the bare DNA molecule, the linking number equals 2720/3.5 = 777. A RecA–DNA filament on the same (length of DNA) has a linking number of 4100/9.5 = 432. The difference in linking number of 432 − 777 = −345 corresponds very well to the experimental values of the negative turns of the magnets, which is 340 ± 10 (N = 5). Our data show that the RecA binding reaction reaches an equilibrium that is force dependent, as shown in Figure 3. The end-to-end distance of the molecule at which the reaction finally stalled varied strongly with different stretching forces. The measured extension corresponded to the length of a partially RecA-coated dsDNA molecule with the remaining uncovered dsDNA in a positively supercoiled state. Because highly supercoiled dsDNA has an end-to-end distance of approximately zero (see Figure 1b), we take the height of the bead as a measure for the length of the partially formed RecA–DNA filament. It should be noted that all torsion is assumed to be absorbed by the remaining free dsDNA; in other words, we assume that the rotational stiffness of a RecA–DNA filament is much larger than that of bare dsDNA. This is a reasonable assumption because Stasiak and Di Capua (4) showed using electron microscopy that the torsion generated during RecA assembly is indeed almost entirely consumed by the remaining DNA in the covalently closed plasmid and that the partially formed RecA–DNA filament hardly consumed any torsion. We now introduce a model that describes the influence of plectoneme formation on RecA assembly on a torsionally constrained DNA molecule. Given an equilibrium situation in which DNA is partitioned into RecA-coated and highly supercoiled portions, the base pairs of the DNA molecule are either consumed in positive plectonemes or covered with RecA. The overall number NDNA of base pairs in the DNA molecule is a constant: NDNA = NRecA + Npl, where NRecA is the number of base pairs covered with RecA and Npl the number of base pairs that is not covered by RecA and which is consumed in plectonemes. As mentioned in the introduction, a single RecA monomer covers 3 bp and therefore NRecA = 3n, where n is the number of bound RecA monomers. The size of a single plectoneme lp is given by lp = 2πR, where R is the radius of a plectoneme (15). The amount of base pairs stored in a single plectoneme is Nlp=(2πR)/(dDNA), in which dDNA is the base pair rise of dsDNA. The heavily supercoiled state exists of multiple positive plectonemes, therefore Npl=mNlp with m the number of plectonemes. Binding of a single RecA monomer (covering 3 bp) induces a change ϕ in the helical pitch of approximately ϕ = 3 (20° − 35°) = −45°. As soon as a torsionally constrained DNA molecule has passed its buckling instability, the torsion induced by RecA is converted into the formation of plectonemes. The number of plectonemes m can therefore be expressed as m=(nϕ−Ωb)/2π, where Ωb is the twist angle at which the DNA molecule starts to buckle, and both ϕ and Ωb are expressed in radians. Before a DNA molecule reaches its buckling instability upon twisting at a given stretching force, its extension remains unchanged and pure twist accumulates (see the data for 3.8 pN in Figure 1b). In this regime, Hooke's law relates the accumulated twist angle Ω and the torsion Γ as follows 1 Γ=CΩL, in which C is the torsional modulus and L the contour length of the molecule. Therefore, the number of base pairs of DNA stored in plectonemes is 2 Npl=mNlp=(nϕ−ΓbLC)×RdDNA. Both the radius of a plectoneme and the buckling instability Γb are force-dependent, given by (16) 3 R=(PkbT2F)α, and 4 Γb=2PkbTF, where P is the persistence length, kb Boltzmann's constant, T the absolute temperature and F the applied stretching force. For the value α, Strick et al. (16) found experimentally a value of 0.4, whereas simple mechanics yields a value of 0.5 (17). In our calculations, we adopt the experimental value of 0.4. Using Equations 2–4, we can determine how many RecA monomers n cover the torsionally constrained DNA molecule with a contour length L (= NDNA dDNA) for a given stretching force F as only the amount of base pairs stored in plectonemes is force-dependent. The overall number of base pairs in the DNA molecule becomes 5 NDNA=(nϕ−ΓbNDNAdDNAC)×RdDNA+3n, which can be rewritten as 6 n=(NDNA+ΓbNDNARC)(3+ϕRdDNA)−1, where the radius of a plectoneme and the buckling instability are given by Equations 3 and 4, respectively. The length of the RecA–DNA filament is LRecA = 3ndRecA, where dRecA is the base pair rise in a RecA–DNA filament, and the amount of DNA consumed in plectonemes is Lpl = Npl dDNA. Up to forces of 3.6 pN in Figure 3, the RecA–DNA filament length increases monotonically with applied stretching force. The solid line in Figure 3 denotes the length of the RecA–DNA filament as derived from the above model using a base pair rise in a RecA–DNA filament of 0.51 nm (3) and a torsional modulus of DNA of 86 kbT·nm (18). The persistence length of the naked DNA molecule (53 ± 2 nm) was obtained from force-extension measurements. Note that the model curve is plotted without any adjustable fitting parameters. The torsional modulus of DNA, however, varies between 75 and 120 kbT·nm depending on the applied analytical method (19). Such a variation shifts the theoretical curve of Figure 3 up or down by only 10%. In this force regime, the proposed model corresponds very well to measured stalling positions. In the model, RecA assembly is stalled by the formation of (positive) plectonemes in the remaining uncovered dsDNA molecule. However, for forces beyond 3.6 pN, a different behavior is observed. The length of the partially formed RecA–DNA filament becomes approximately constant at a low value of about LRecA = 0.54 μm, much lower than the predicted values from the proposed model. We can understand this as follows. In order to create the first plectoneme, the buckling instability of the DNA molecule has to be overcome. Whether this is possible or not depends on the torsion produced by the partially formed RecA–DNA filament. To determine this torsion using Equation 1, the twist angle and the length in which the twist accumulates have to be deduced. As stated in the model, every RecA monomer covers 3 bp and induces a change in the helical pitch of approximately ϕ = −45°. The DNA molecule that is partially covered by n RecA monomers is, therefore, unwound by Ω=nϕ=LRecA/3dRecAϕ=44 rotations, which is consumed by the remaining uncovered DNA molecule with a length Lfree DNA given by Lfree DNA=L−11.5×LRecA=2.72−11.5×0.54=2.36 μm. The torsion produced by the partially formed RecA–DNA filament on the remaining uncovered DNA is therefore calculated from Equation 1 using a torsional modulus of DNA of 86 kbT·nm (18) to equal 41 ± 3 pN·nm or about 10.1 ± 0.7 kbT. This is indeed equal to the value of 9.6 ± 0.2 kbT needed to overcome the buckling instability at stretching forces up to 3.6 pN (Equation 4). Our calculations thus confirm that RecA assembly on a torsionally constrained DNA molecule generates an amount of torsion that is insufficient to overcome the buckling instability at forces above 3.6 pN. Using a similar setup, Fulconis et al. (6) demonstrated that applying an external torque on a RecA-covered DNA molecule forced the filament to depolymerize. In the presence of ATP, they estimated a RecA depolymerization torque between 3 and 5 kbT. This value, however, appears to be a lower bound, as the buckling force in their experiments is between 1 and 2 pN, which, using Equation 4, yields a RecA depolymerization torque between 5 and 7 kbT. Also, the method by Fulconis et al. (6) to determine the depolymerization torque is rather sensitive to fluctuations in the measured value of the slope of their rotation curves; a change in this measured value of 10% causes the obtained RecA depolymerization torque to change by 100%. The energy that a RecA monomer gains for binding to dsDNA is used to unwind and stretch the dsDNA molecule. The energy E required to locally unwind dsDNA by a single RecA monomer is E ≅ Γϕ, as the induced torsion in the remaining uncovered DNA remains approximately constant. The maximum built-up torsion in dsDNA during RecA assembly is Γ = 10.1 kbT (as calculated above) and the change in helical pitch is ϕ=45°×2π/360°=π/4, which together yield a torsional energy of E = 7.9 ± 0.5 kbT. The elongation of DNA by a single RecA monomer reduces the torsional energy by 3 (dRecA − dDNA) F, which is 0.1–1.5 kbT for F = 1–10 pN due to the work carried out by the applied stretching force. The energy to unwind and stretch DNA by a single RecA monomer is therefore about 7 kbT. This provides a lower bound for the binding energy of RecA. Using biochemical assays, the binding energy of a RecA monomer to ssDNA has been determined to be 10 kbT (20), whereas in the experiments of Fulconis et al. (6) a binding energy of RecA to dsDNA of ∼6 kbT was obtained. In the presence of ATP, RecA is able to bind and dissociate from dsDNA owing to ATP hydrolysis. RecA disassociation is blocked in the presence of ATPγS, a non-hydrolyzable ATP analog (21). Applying an external torque depolymerizes filaments assembled in the presence of ATPγS (6). In this case, the external torque at which the RecA–DNA filament starts to depolymerize is, therefore, a measure for the unbinding energy of ATPγS bound RecA from dsDNA. The question arises whether the binding energy is equivalent to the unbinding energy. In the presence of ATPγS, an external torque of more than 13 kbT had to be applied to dissociate RecA from dsDNA giving a lower bound for the unbinding energy of 10 kbT (22). This value is comparable with the energy that is available upon ATP hydrolysis of 12.3 kbT. When exposed to high torque (Γ ≳ 8 kbT) DNA can undergo a structural transition from regular B-DNA into supercoiled P-DNA (Pauling-like structure with exposed bases) (23). As RecA assembly creates up to 10 kbT of torsion in the remaining DNA molecule, the formation of supercoiled P-DNA could hinder further RecA binding. However, this structural transition is highly dependent on the ionic conditions in the flow cell. Addition of 5 mM MgCl2 shifts the critical torque at which the structural transition occurs to 13 kbT (22). Owing to the presence of 10 mM MgCl2 in our experiments, it is therefore unlikely that supercoiled P-DNA has formed. We consider two possible biological roles for the DNA torsion induced by RecA filament formation on dsDNA. RecA filaments on dsDNA occur as a result of strand invasion and joint molecule formation in homologous recombination. Torsion introduced here in the target DNA molecule, usually the chromosome, could function to displace the E.coli proteins responsible for chromosome condensation in order to complete the recombination process. Chromosomal DNA in E.coli is condensed by architectural proteins, including HU, H-NS, IHF and Fis (24). Though several studies report force spectroscopy analysis of these proteins bound to DNA (7,25), none has addressed a critical torque, as described here for RecA. Such critical torque values would be interesting for comparison with RecA to determine whether the architectural proteins could be displaced. RecA filaments also form on dsDNA nucleated at sites of damage, such as ultraviolet (UV) lesions. Under normal growth conditions, bacterial replication forks are stalled by DNA damage (26,27), such as UV-irradiation-induced lesions (28). The repair of such stalled replication forks is a significant function of recombination systems (29). One mechanism for rescuing stalled replication involves fork reversal resulting in a Holliday junction structure called a ‘chicken foot’. Postow et al. (30) have shown that positive torsional strain is absorbed by a regression of the replication fork and calculated that this would require very little energy, only 5 kcal/mol (8.4 kbT). Formation of RecA filaments at sites of DNA damage ahead of the replication fork could generate up to 4.1 kcal/mol (7 kbT) of positive torsional energy, which would be sufficient to drive the fork reversal and is in accordance with the value estimated by Postow et al. (30). CONCLUSION In this paper, we quantitatively described the binding of RecA to torsionally constrained dsDNA. Upon binding, a RecA monomer both stretches and twists the backbone of the DNA molecule. The built-up of torque is stored either in writhe (plectonemes) or in pure twist depending on the applied force. A torsional constraint of the DNA molecule leads to stalling of RecA assembly because the remaining uncovered DNA cannot absorb more torsion. Rotation of the magnets results in the expected elongation of 50%. The amount of magnet turns is a measure for the change in linking number accompanied with the binding of RecA. The assembly reaction stalls whenever the extra torsion stored in the remaining free dsDNA reaches a value of 10.1 kbT. At this amount of torsion, it is not energetically favorable for RecA to bind to its substrate. We have discussed the possible biological role of RecA-induced torsion in regression of replication forks and displacement of condensing proteins.